|Publication number||US7460565 B2|
|Application number||US 10/695,069|
|Publication date||Dec 2, 2008|
|Filing date||Oct 28, 2003|
|Priority date||Oct 28, 1999|
|Also published as||US6674772, US20040086002|
|Publication number||10695069, 695069, US 7460565 B2, US 7460565B2, US-B2-7460565, US7460565 B2, US7460565B2|
|Inventors||William J. Dally, John W. Poulton|
|Original Assignee||Rambus Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (33), Non-Patent Citations (6), Referenced by (7), Classifications (13), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a Divisional of application Ser. No. 09/557,564, filed on Apr. 25, 2000 now U.S. Pat. No. 6,674,772 which also claims the benefit of U.S. Provisional Application No. 60/161,977, filed Oct. 28, 1999, the entire teachings of which are incorporated herein by reference.
High-speed data communication systems commonly use multiplexing transmitters and demultiplexing receivers. In such a system, as illustrated in
Phase-locked loop clock multipliers have been used to multiply the frequency of the reference clock to generate the bit clock. In
An alternative prior art multiplexing data communication system that uses a multi-phase clock rather than a clock multiplier is illustrated in
Prior art data communication systems based on PLL clock multipliers and multi-phase DLLs have large amounts of jitter due to the method used to generate timing signals. Phase-locked loop based clock multipliers have large amounts of jitter because the phase error at the end of each cycle accumulates until the control loop can respond. As described in Kim, Weigant, and Gray, “PLL/DLL System Noise Analysis for Low Jitter Clock Synthesizer Design,” ISCAS 1994, pp. 31-38, this error accumulation multiplies the jitter of the basic delay elements by a factor that is inversely proportional to the loop bandwidth. For typical phase-locked loops, the jitter is multiplied by a factor of at least 10.
Communication systems based on multi-phase delay-locked loops do not accumulate jitter from cycle-to-cycle like PLL clock multipliers. However, they do introduce jitter due to cumulative phase mismatches. Due to device mismatches in the delay elements, there is a variation in the delay of each stage of the delay line. These phase mismatches accumulate over the length of the delay line leading to large jitter values, particularly when the multiplexing rate is high.
To reduce the number of precision clock signals required in a multiplexing circuit, either a multiplexer or a demultiplexer, a multiplexing circuit in a data communication circuit is formed in multiple stages. Specifically, a data communication circuit for serializing data on a communication link comprises a higher frequency data multiplexing stage and a lower frequency data multiplexing stage. The higher frequency data multiplexing stage is coupled between the communication link and an intermediate frequency signal. The lower frequency data multiplexing stage is coupled to the higher frequency data multiplexing stage between the intermediate frequency signal and the lower frequency signal. A data stream on the communication link has a first number of bitstreams, typically a single bitstream, and the intermediate frequency signal has a data stream of a second number of parallel bits greater than the first number. The lower frequency signal has a data stream of a third number of parallel bits greater than the second number. The stages are clocked by a clock signal from a common clock source, but the clock signal is more precisely distributed to the higher frequency stage.
In a data transmitter, each of the data multiplexing stages is a data multiplexer, and in a data receiver, each of the data multiplexing stages is a data demultiplexer.
In a multiplexer, a lower frequency data multiplexer multiplexes parallel bits of a lower frequency data signal to an intermediate frequency signal of parallel bits. A higher frequency data multiplexer multiplexes the bits of the intermediate frequency signal to a higher frequency signal on the communication link.
In a demultiplexer, a higher frequency data demultiplexer demultiplexes the data on the communication link to an intermediate frequency signal. A lower frequency data demultiplexer further demultiplexes the intermediate frequency signal.
Preferably, the entire multiplexing circuit is on a single chip. The common clock signal may be divided by a ring counter for application to the lower frequency stage. The preferred clock is a multiplying delay locked loop bit clock generator.
In one embodiment, the data stream on the communication link is a one-bit-wide bitstream, and the intermediate frequency signal comprises two parallel bits. In another embodiment, the second number of parallel bits is greater than 2. In the preferred implementation of that embodiment, the higher frequency data multiplexing stage is clocked by an N-phase overlapping clock. Each bit of the intermediate frequency signal is enabled in the higher frequency data multiplexing stage by concurrence of two clock phases.
The foregoing and other objects, features and advantages of the invention will be apparent from the following more particular description of preferred embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention.
A description of preferred embodiments of the invention follows.
A clock-multiplying delay-locked loop previously used as a microprocessor clock is illustrated in
In operation, control block 145 sets multiplexer control line 146 to select the reference clock as input to the delay line. Then, after a rising edge on the reference clock, control block 145 switches line 146 to recirculate the bit clock 112, effectively forming the ring oscillator. After N−1 rising edges on the bit clock, control switches back to select the reference clock to apply a single rising edge to the delay line. The delay flip flop 143 serves as a phase comparator to provide a delay control signal vctrl through a charge pump 144. Adjustment of the delay serves to phase synchronize blck with the rising edge of rclk, the frequency of bclk being a multiple of the frequency of rclk.
Multiplying delay-locked loops have been used to multiply and deskew clocks in microprocessors as described in A Waizman, “A Delay Line Loop for Frequency Synthesis of De-Skewed Clock,” IEEE International Solid-State Circuits Conference, 1994, pp 298-299.
Multiplying DLLs have not been used in data communication systems to date. A conventional multiplying DLL is not well-suited to data communications due to the fact that it has high jitter, the phase difference between earliest and latest edges of a cycle, due to two factors. First, the flip-flop 144 is a ‘bang-bang’ phase comparator that leads to considerable jitter when the loop is locked. This jitter is due to dithering of the control voltage to the delay element about the correct voltage. Second, and most importantly, because the prior multiplying DLL does not precisely align the rising edges of the bit clock and the reference clock, there is a considerable phase offset that appears as jitter during the cycle that the clock multiplexer switches to select the reference clock as input (last cycle mismatch).
The waveforms of
The waveforms of
The present invention overcomes the high jitter of prior art data communication timing circuits by using a multiplying DLL with a low-offset, proportional phase comparator to eliminate both the dither and the ‘last-cycle’ mismatch. A block diagram of the present invention is shown in
A block diagram of a data transmission system of the present invention using the multiplying delay-locked loop is shown in
Proportional phase comparator 147 directly compares the phase of reference clock rclk 111 and bit clock bclk 112 and generates signals up and down which are proportional to the amount of phase difference. These signals cause charge pump 144 to adjust the level of vctrl which in turn controls the delay of the delay line. In the preferred embodiment this is done by using the select signal 146 as a window signal to a combined phase comparator and charge pump of the type described in pending patent application Ser. No. 09/414,761. Signals rclk and bclk are only examined by the phase comparator when window signal sel 146 is high. When rclk rises before bclk during the timing window when sel is asserted, the delay line is too slow. Thus, the phase comparator causes signal up to be asserted when rclk=1, bclk=0, and sel=1. This causes vctrl to be increased, speeding up the delay line and bringing rclk and bclk into alignment. Similarly, if bclk rises before rclk when sel is asserted, the delay line is too fast. To correct for this case, the phase comparator asserts signal down when rclk=0, bclk=1, and sel=1. This causes vctrl to be reduced, slowing the delay line and bringing rclk and bclk into alignment.
One skilled in the art will understand that the present invention can be realized with other phase comparators. For example, one may use a phase comparator that is not combined with a charge pump. Also, one may use a flip-flop phase comparator as in
To generate a bit clock with very low jitter, it is important that phase comparator 147 directly compare rclk 111 with bclk 112 and not with dclk, the output of a divider, as in
Detail of control logic 145 that generates multiplexer select signal 146 is illustrated in
In the preferred embodiment, delay line 142 is realized with a differential inverter delay line controlled by a current regulator as described in pending patent application Ser. No. 09/453,368 filed Dec. 7, 1999 by Dally et al. for Low-Power Low-Jitter Variable Delay Timing Circuit. A block diagram of the delay line is shown in
A block diagram of an individual delay element 151 is shown in
Waveforms illustrating operation of the multiplying DLL of the present invention when the loop has converged are shown in
The waveforms shown in
Because sel is high from the fourth falling edge of bclk to the time rclk rises, the multiplexer selects rclk as input to the delay line during this period. The bclk pulse is stretched until rclk rises which causes sel to fall, x to rise, and bclk to fall after the delay of the delay line. The net result is that when the delay line is too fast, every fourth pulse on bclk is stretched. This stretching allows bclk to catch up with the reference clock every rclk cycle. Thus, no phase error is accumulated from one rclk cycle to the next.
The circuit will operate properly regardless of how fast the delay line is (as long as the divide by N counter and the select flip-flop can keep up with the fast bit clock). The down pulse will simply last from the fourth rising edge of bclk to the rising edge of rclk. Note that if the delay line is very fast, last may rise before rclk falls. In this case, the logic of
If the delay line is very slow, rclk may go high before sel is asserted. In this case the up pulse will last the entire low-time of bclk. When sel switches the multiplexer to select an already high rclk, x goes high almost immediately. This results in a very short low pulse on x and a very short high pulse on bclk. More importantly, because this falling edge of bclk is caused by sel rising, not rclk rising, phase error does accumulate in this case. When the delay line is more than a half a bit clock slow, the bit clock will slip farther and farther behind rclk until it slips to the point where rclk is low. At this point the circuit sees rclk as being late and asserts down until rclk rises. The multiplying delay-locked loop will still converge from this state as long as the charge transferred by the one down pulse is smaller than the charge transferred by the series of up pulses. If necessary, convergence can be forced by inhibiting the first down pulse after a series of two or more up pulses.
One skilled in the art will understand that the frequency-multiplying DLL clock generator described here may be realized in many different ways. The circuit may have a different multiplication rate from that shown here or may have a programmable multiplication rate. The circuit may use a variety of delay elements including source-coupled logic elements, current-starved elements, and single-ended CMOS gates. The circuit may employ many different types of multiplexers including pass-gate multiplexers and static CMOS multiplexers. Different phase comparators can be employed including flip-flop and exclusive-OR phase comparators. Different types of charge pumps and other loop filters may be used in the circuit. Also, the control logic to control the multiplexer can be realized in a number of different ways including static CMOS logic, dynamic CMOS logic, and source-coupled FET logic.
Another problem with prior art data communication systems is that, at high multiplexing rates (e.g., 10:1), there is a high clock fan out, and hence a high clock load in both the data multiplexer and data demultiplexer. This high clock load leads to jitter in the output signal because it is difficult to fan out the clock signal to all of the loads with balanced delay. It also results in increased power dissipation.
The illustrated system divides the multiplexer into two or more stages. This reduces jitter by reducing the number of locations that the bit clock must be distributed with precise delay.
Prior art data communication systems using DLLs to generate timing signals typically generated a number of phases equal to the multiplication rate between the reference clock and the data rate. Distributing all of these clock phases with balanced delay is a challenging problem and mismatches in distribution usually account for considerable jitter in the system. For example, Dally and Poulton, “Transmitter Equalization for 4 Gb/s Signaling,” IEEE Micro, January-February 1997, pp. 48-56, describes such a DLL-based data communication system which uses a 10-phase clock in the transmitter and a 20-phase clock in the receiver. Each of these 30-phases must be distributed to all of the clock loads with balanced delay. In this system, the 10:1 multiplexer is performed in a single stage. Even in systems where several stages of logic are used to implement such multiplexers, the multiplexing is logically done in a single stage. In such systems, the data is reduced from an N-wide bus at the reference clock rate to a single-bit line at N times the reference clock rate using a single set of precision timing signals.
The present invention of a low-jitter frequency-multiplying DLL enables the design of data communication systems that have the advantages of DLL timing circuits (no accumulation of jitter from one reference clock cycle to the next) and at the same time use high-frequency clocks. One use of such high-frequency clocks is to perform the multiplexing and demultiplexing in multiple stages. By operating the last stage of the multiplexer and the first stage of the demultiplexer at the highest possible frequency, the number of clock phases that must be distributed and the number of clock loads that require a precision clock are reduced. This results in lower overall system jitter.
A two-stage data multiplexer 102 of
This circuit has lower jitter than prior art one-stage multiplexers because the bit clock must only drive the two FETs 156 and 157 of the final multiplexer with balanced fan out. All other clock loads in this circuit are non-critical. In a prior art one-stage multiplexer, the bit clock must drive eight multiplexer loads with balanced fan out.
The multiplexer 102 operates off of a bit clock, bclk 117 that runs at half the bit rate. Two bits are transmitted on each cycle of bclk, one when bclk is high, and the second when bclk is low. Multiplexer 102 of
A four-bit ring counter formed by flip-flops 168-171 sequences the first stage of the multiplexer. These flip-flops are initialized to state 1000 (e1=1 and e2=e3=e4=0) and the state rotates left on each rising edge of bclk. Thus during the first cycle, e1 172 is asserted causing FET 164 to turn on and in turn enabling the top two bits of input 113 onto signal 176. During the second cycle, the state is 0100 with e2 173 asserted, FET 165 on, and the second pair of bits from input 113 enabled onto output 176. During the third and fourth cycles, this pattern continues with the third and fourth pairs being enabled onto signal 176. The pattern then repeats starting over with e1 being asserted. During each cycle, the bit pair enabled onto signal 176 is captured in latch 163 which is gated by bclkN.
The clocking of flip-flops 168-171 that make up the ring counter which generates enables e1 through e3, and the clocking of latch 163 which samples the selected bit pair need not be very precise. As long as these clocks remain within a half-clock period centered about the optimum timing point, the output of latch 163 will remain stable during bclkP high when it is sampled by the second stage of the multiplexer. The relaxed timing of the clock to the first stage multiplexer allows a less precise buffered version of bclk to be used to drive the clock loads in this stage and simplifies the distribution of this clock.
The second stage of the multiplexer 162 is a 2:1 multiplexer controlled by the complementary phases of bclkP and bclkN and consisting of latch 158 and FETs 156 and 157. During the first half cycle (bclkP asserted), the high bit of signal 177 is enabled to output 178 via FET 156. At the same time, the low bit of signal 177 is sampled by latch 158 which holds this value to be output on the second half cycle. During the second half cycle (bclkN asserted), FET 157 enables the contents of latch 158, which holds the low bit of signal 177, to output 178. The relative timing of signals bclkP and bclkN and their distribution as low precision signals 301 and 303 through devices 252 and 253 is not critical. On the other hand, the relative timing of signals bclkP and bclkN and their distribution to FETs 156 and 157 as signals 302 and 304 through drivers 250 and 251 must be carefully controlled as any timing errors in the signals on the gates of these FETs translates directly to jitter on the output signal. However, this timing and distribution is relatively simple as the precision clock need only be distributed to two clock loads, FETs 156 and 157.
Waveforms showing operation of the two-stage multiplexer are illustrated in
With a multi-stage multiplexer, it is advantageous to keep the number of inputs to the final stage of multiplexing as small as possible because this minimizes the number of clock loads for the precision clock. A multi-stage multiplexer with a 2:1 second stage, as in
A block diagram of a second stage multiplexer 162 that performs 4:1 multiplexing from a quarter-bit-rate bit clock, bclk 117P and 117N, is shown in
The 4:1 multiplexer of
Four bit data enters the 4:1 second-stage multiplexer of
At very-high bit rates the bandwidth of multiplexer output signal 178 and output driver 104 become critical. These signals may toggle as frequently as every 1.5 FO4 delays. To improve the bandwidth of signal 178, the preferred embodiment adds a grounded PFET load resistor 254 to this node that limits the swing of this signal and boosts its bandwidth. The preferred embodiment also uses a source-coupled FET output driver 104 to realize adequate bandwidth.
One skilled in the art will understand that a multi-stage multiplexer can be realized in a number of different ways. There may be applications where it is advantageous to use more than two stages. Three or more stages may be appropriate at very high multiplexing rates. Different multiplexing rates may be used and a multiplexer may be constructed with a programmable rate. The enable signals for the first stage may be generated in different ways, for example using a counter. The pass-gate multiplexers shown in this design may be replaced by other types of multiplexers, for example static CMOS gate-based multiplexers. Also, the latches and flip-flops in these multiplexers may be replaced by other types of clocked storage elements.
The timing of a demultiplexer can also be improved by splitting it into two or more stages. By operating the first stage with the lowest possible demultiplexing rate the number of precise clock phases that are required and the number of clock loads that must be driven with balanced delay are both minimized, just as with the multi-stage multiplexer. As with the multiplexer, the use of multi-stage demultiplexers in systems with DLL timing is enabled by the use of the frequency-multiplying DLL of the present invention.
A block diagram of a two-stage 1:8 demultiplexer is shown in
Data enters the second stage of the multiplexer on two-bit line 203. This data is aligned with the rising edge of bclk by two-bit latch 206. Signal 207 on the output of this latch is then distributed to four two-bit wide latches 199-204 that are gated by strobes s1-s4 191-194. The strobes are generated by a ring counter formed by flip-flops 195-198 in a manner similar to that used in the multiplexer. The four strobes sample two-bit values from signal 207 into the four latches in sequence. Strobe s1 samples the first two bits into latch 199 on the first bclk. On the next clock, strobe s2 samples the second two bits into latch 200, and so on. The outputs of this multiplexer are each valid for one reference clock time, but staggered across the four pairs. A latch gated with s4 may be used to align the four output pairs.
Waveforms showing the operation of the two-stage demultiplexer are shown in
As with last-stage multiplexers, first-stage demultiplexers must operate off of a clock with a period of at least 6 FO4 delays. Thus, with bit periods shorter than 3 FO4delays the 1:2 first-stage demultiplexer operating off of a half-bit-rate clock cannot keep up. The advantages of multi-stage demultiplexers can still be realized in this case by operating with a clock at a smaller fraction of the bit rate and a corresponding higher degree of demultiplexing. Such a design still has fewer critical clock phases and critical clock loads than conventional DLL-based demultiplexers that use a number of phases equal to the multiplication rate between the reference clock and the bit rate.
A first-stage demultiplexer 190 with a 1:4 demultiplexing rate is shown in
This 1:4 first-stage demultiplexer is intended to be used in conjunction with a second-stage demultiplexer such as the second stage of the two-stage demultiplexer of
One skilled in the art will understand that a multi-stage demultiplexer can be realized in a number of different ways. The number of stages may be greater than two, and each stage may have a different demultiplexing rate or a programmable demultiplexing rate. The first stage may employ separate amplifiers and flip-flops rather than combined clocked receive amplifiers. Different circuits may be used to generate the strobes for the second stage of the demultiplexer. Also, the latches and flip-flops in the demultiplexer may be replaced by other types of clocked storage elements.
While this invention has been particularly shown and described with references to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4387459 *||Jun 22, 1981||Jun 7, 1983||Rockwell International Corporation||Non-buffered loop-timed channel bank|
|US4626796||Mar 1, 1985||Dec 2, 1986||General Electric Company||Digital apparatus and method for programmably phase shifting an audio tone|
|US5260608||Jan 30, 1991||Nov 9, 1993||Bull, S.A.||Phase-locked loop and resulting frequency multiplier|
|US5278702||Apr 12, 1991||Jan 11, 1994||Western Digital Corporation||Data synchronizer with symmetric window generation|
|US5361254||Nov 30, 1992||Nov 1, 1994||Siemens Aktiengesellschaft||Switch mechanism for switching signals at inputs onto outputs and switching networks for interlinking first and second transmission media|
|US5432481||Nov 17, 1994||Jul 11, 1995||Kabushiki Kaisha Toshiba||Phase-locked loop circuit|
|US5514990||Dec 23, 1994||May 7, 1996||Kabushiki Kaisha Toshiba||Frequency multiplier circuit|
|US5537069||Mar 30, 1995||Jul 16, 1996||Intel Corporation||Apparatus and method for selecting a tap range in a digital delay line|
|US5646564||Apr 12, 1996||Jul 8, 1997||Xilinx, Inc.||Phase-locked delay loop for clock correction|
|US5675584 *||Dec 29, 1995||Oct 7, 1997||Sun Microsystems, Inc.||High speed serial link for fully duplexed data communication|
|US5717353||Sep 27, 1995||Feb 10, 1998||Kabushiki Kaisha Toshiba||Clock signal generating circuit|
|US5724361 *||Mar 12, 1996||Mar 3, 1998||Lsi Logic Corporation||High performance n:1 multiplexer with overlap control of multi-phase clocks|
|US5786715||Jun 21, 1996||Jul 28, 1998||Sun Microsystems, Inc.||Programmable digital frequency multiplier|
|US5787132 *||Jul 20, 1995||Jul 28, 1998||Nippondenso Co., Ltd.||Data communication system having improved synchronization capability|
|US6014177||Nov 26, 1997||Jan 11, 2000||Nec Corporation||Video display apparatus having phase-locked loop used for synchronizing a horizontal scan frequency with a synchronizing input signal frequency|
|US6028462 *||Aug 22, 1997||Feb 22, 2000||Lsi Logic Corporation||Tunable delay for very high speed|
|US6037812||May 18, 1998||Mar 14, 2000||National Semiconductor Corporation||Delay locked loop (DLL) based clock synthesis|
|US6043717||Sep 22, 1998||Mar 28, 2000||Intel Corporation||Signal synchronization and frequency synthesis system configurable as PLL or DLL|
|US6075406||Jun 16, 1998||Jun 13, 2000||Macronix International Ltd.||High speed differential charge pump apparatus and method using a replica circuit to control common mode output voltage|
|US6087864||Nov 10, 1997||Jul 11, 2000||Kabushiki Kaisha Toshiba||Digital frequency multiplier circuit including delay circuit|
|US6097777||Jun 20, 1997||Aug 1, 2000||Pioneer Electronic Corporation||Phase locked loop circuit|
|US6114915||Mar 23, 1999||Sep 5, 2000||Altera Corporation||Programmable wide-range frequency synthesizer|
|US6208183||Apr 30, 1999||Mar 27, 2001||Conexant Systems, Inc.||Gated delay-locked loop for clock generation applications|
|US6211742||Nov 3, 1999||Apr 3, 2001||Broadcom Corporation||Lock detector for phase locked loops|
|US6240274 *||Apr 21, 1999||May 29, 2001||Hrl Laboratories, Llc||High-speed broadband wireless communication system architecture|
|US6259755||Mar 30, 1998||Jul 10, 2001||Nec Corporation||Data clock recovery PLL circuit using a windowed phase comparator|
|US6269088 *||Aug 1, 1996||Jul 31, 2001||Hitachi, Ltd.||CDMA mobile communication system and communication method|
|US6275072||Oct 7, 1999||Aug 14, 2001||Velio Communications, Inc.||Combined phase comparator and charge pump circuit|
|US6285225||Jul 23, 1999||Sep 4, 2001||Samsung Electronics Co., Ltd.||Delay locked loop circuits and methods of operation thereof|
|US6310895 *||Mar 27, 1998||Oct 30, 2001||Telefonaktiebolaget Lm Ericsson (Publ)||Clock selector system|
|US6310913 *||May 20, 1997||Oct 30, 2001||Asahi Kasei Kabushiki Kaisha||Circuit and method for modulating pulse width|
|US7161937 *||Dec 13, 1996||Jan 9, 2007||Intel Corporation||Method and apparatus for routing encoded signals through a network|
|US20010015663||May 4, 2001||Aug 23, 2001||Velio Communications, Inc.||Combined phase comparator and charge pump circuit|
|1||"Engineering and Operations in the Bell System," Second Edition, R.F. Rey, Technical Editor, AT&T Bell Laboratories, Murray Hill, N.J., 1982-1983, pp. 386-393.|
|2||Dally, William J. and Poulton, John W., "Digital Systems Engineering," Cambridge University Press, pp. 459-460, 615-620 and 626-627.|
|3||Dally, William J. and Poulton, John W., "Transmitter Equalization for 4Gb/s Signaling," IEEE Micro, Jan.-Feb. 1997, pp. 48-56.|
|4||Dally, William J. and Poulton, John W., Digital Systems Engineering, Cambridge University Press, 1998, pp. 428-447, 537-540 and 547-548.|
|5||Kim, Weigant and Gray, "PLL/DLL System Noise Analysis for Low Jitter Clock Synthesizer Design," ISCAS, 1994, pp. 31-34.|
|6||Waizman, A., "A Delay Line Loop for Frequency Synthesis of De-Skewed Clock," IEEE International Solid-State Circuits Conference, 1994, pp. 298-299.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8014485||May 17, 2007||Sep 6, 2011||Advanced Micro Devices, Inc.||Techniques for integrated circuit clock management using multiple clock generators|
|US8195849||Nov 6, 2009||Jun 5, 2012||Advanced Micro Devices, Inc.||Device and method for transferring data between devices|
|US8575972 *||Mar 23, 2009||Nov 5, 2013||Advanced Micro Devices, Inc.||Digital frequency synthesizer device and method thereof|
|US20080285696 *||May 17, 2007||Nov 20, 2008||Advanced Micro Devices, Inc.||Techniques for integrated circuit clock management using multiple clock generators|
|US20100049887 *||Feb 25, 2010||Advanced Micro Devices, Inc.||Device and method for transferring data between devices|
|US20100237924 *||Sep 23, 2010||Advanced Micro Devices, Inc.||Digital frequency synthesizer device and method thereof|
|WO2015142350A1 *||Mar 21, 2014||Sep 24, 2015||Empire Technology Development Llc||Bandwidth amplification using pre-clocking|
|U.S. Classification||370/544, 327/415|
|International Classification||H03K5/00, H03L7/081, H04J3/06, H04J3/22, H04J3/04|
|Cooperative Classification||H04J3/047, H03L7/0812, H03K2005/00032, H04J3/0685, H03K2005/00234|
|Dec 29, 2003||AS||Assignment|
Owner name: RAMBUS, INC., CALIFORNIA
Free format text: ACQUISITION OF ASSETS;ASSIGNOR:VELIO COMMUNICATIONS, INC.;REEL/FRAME:014221/0988
Effective date: 20031224
|Feb 17, 2004||AS||Assignment|
Owner name: RAMBUS INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:VELIO COMMUNIATIONS, INC.;REEL/FRAME:014972/0386
Effective date: 20031224
|Feb 17, 2009||CC||Certificate of correction|
|Jun 4, 2012||FPAY||Fee payment|
Year of fee payment: 4